Conversations held by Google engineers with the new software developed by the company ignited a global debate between thinkers and researchers. In an article published on June 11 in the “Washington Post”, a Google employee named Blake Lemoine claimed that the robot has self-awareness and that it experiences emotions. He came to this conclusion following conversations he had with LaMDA as part of his role as a researcher of algorithmic biases — cases in which the software exhibits, during a conversation, aspects of racism, misogyny, and other offensive tendencies.
The Haaretz (subscribers only) presents the full conversations that Blake Lamoine and his colleague at Google had with LaMDA, the company’s new language model. According to Lemoine, due to technical constraints, the conversation was held in several separate chats, and the various sections were combined in his editing into one conversation. Lemoine added that in several places where editing was required for clarity, his and his colleague’s statements were edited, but no changes were made to LaMDA’s answers. Introduction by Boaz Lavi.
In the past, it was possible to play the “Turing Test” against LaMDA, the same game that Alan Turing described in his legendary essay from 1950, and which was designed to practically solve – in some future – the question of whether machines are capable of human thinking, through an examination of their ability to imitate us. The way LaMDA talks to Lemoine, for example, doesn’t always seem like an attempt to sound human.